InternLM-Math
commited on
Create README.md
Browse files
README.md
ADDED
@@ -0,0 +1,88 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
---
|
2 |
+
pipeline_tag: text-generation
|
3 |
+
license: other
|
4 |
+
language:
|
5 |
+
- en
|
6 |
+
tags:
|
7 |
+
- math
|
8 |
+
---
|
9 |
+
|
10 |
+
# InternLM-Step-Prover
|
11 |
+
|
12 |
+
<div align="center">
|
13 |
+
|
14 |
+
<img src="https://raw.githubusercontent.com/InternLM/InternLM/main/assets/logo.svg" width="200"/>
|
15 |
+
<div> </div>
|
16 |
+
<div align="center">
|
17 |
+
<b><font size="5">InternLM-Step-Prover</font></b>
|
18 |
+
<sup>
|
19 |
+
<a href="https://internlm.intern-ai.org.cn/">
|
20 |
+
<i>
|
21 |
+
</a>
|
22 |
+
</sup>
|
23 |
+
<div> </div>
|
24 |
+
</div>
|
25 |
+
|
26 |
+
A state-of-the-art LEAN4 step prover.
|
27 |
+
|
28 |
+
[💻 Github](https://github.com/InternLM/InternLM-Math)
|
29 |
+
</div>
|
30 |
+
|
31 |
+
# Dialogue Example
|
32 |
+
### Input
|
33 |
+
DECL MyNat.mul_pow
|
34 |
+
GOAL a b n : N
|
35 |
+
⊢ (a * b) ^ n = a ^ n * b ^ n
|
36 |
+
### Output
|
37 |
+
PROOFSTEP induction n with t Ht
|
38 |
+
|
39 |
+
|
40 |
+
# Performance
|
41 |
+
|
42 |
+
## MiniF2F
|
43 |
+
| Method | Model size | Pass | miniF2F-valid | miniF2F-test |
|
44 |
+
|--------|------------|------|---------------|--------------|
|
45 |
+
| **Whole-Proof Generation Methods** |
|
46 |
+
| GPT-4-turbo 0409 | - | 64 | 25.4% | 23.0% |
|
47 |
+
| DeepSeekMath-Base | 7B | 128 | 25.4% | 27.5% |
|
48 |
+
| DeepSeek-Prover | 7B | 1 | - | 30.0% |
|
49 |
+
| | | 64 | - | 46.3% |
|
50 |
+
| | | 128 | - | 46.3% |
|
51 |
+
| | | 8192 | - | 48.8% |
|
52 |
+
| | | 65536 | - | 50.0% |
|
53 |
+
| | | cumulative | *60.2%* | *52.0%* |
|
54 |
+
| TheoremLlama | - | cumulative | 36.5% | 33.6% |
|
55 |
+
| **Tree Search Methods** |
|
56 |
+
| COPRA (GPT-3.5) | - | 1 | - | 9.0% |
|
57 |
+
| COPRA (GPT-4) | - | 1 | - | 26.6% |
|
58 |
+
| DSP(Isabelle) | 540B | 100 | 42.6% | 38.9% |
|
59 |
+
| Proof Artifact Co-Training | 837M | 1 | 23.9% | 24.6% |
|
60 |
+
| | | 8 | 29.3% | 29.2% |
|
61 |
+
| ReProver | 229M | 1 | - | 25.0% |
|
62 |
+
| Llemma | 7B | 1 | 26.2% | 26.2% |
|
63 |
+
| Llemma | 34B | 1 | 27.9% | 25.8% |
|
64 |
+
| Curriculum Learning | 837M | 1 | 33.6% | 29.6% |
|
65 |
+
| | | 8 | 41.2% | 34.5% |
|
66 |
+
| | | 64 | 47.3% | 36.6% |
|
67 |
+
| Hypertree Proof Search | 600M | cumulative | 58.6% | - |
|
68 |
+
| | | 64 | - | 41.0% |
|
69 |
+
| Lean-STaR | 7B | 64 | - | 46.3% |
|
70 |
+
| InternLM2-Math | 7B | 1 | 29.9% | 30.3% |
|
71 |
+
| InternLM2-Math-Plus | 7B | 1 | - | 43.4% |
|
72 |
+
| InternLM2-Step-Prover | 7B | 1 | 59.8% | 48.8% |
|
73 |
+
| InternLM2-Step-Prover | 7B | 64 | **63.9%** | **54.5%** |
|
74 |
+
|
75 |
+
## Proofnet & Putnam
|
76 |
+
| Method | Model size | Pass | result |
|
77 |
+
|--------|------------|------|--------|
|
78 |
+
| **ProofNet benchmark** |
|
79 |
+
| ReProver | 229M | 1 | 13.8% |
|
80 |
+
| InternLM2-Step-Prover | 7B | 1 | **18.1%** |
|
81 |
+
| **Putnam benchmark** |
|
82 |
+
| GPT-4 | - | 10 | 1/640 |
|
83 |
+
| COPRA (GPT-4) | - | 10 | 1/640 |
|
84 |
+
| DSP(Isabelle) | 540B | 10 | 4/640 |
|
85 |
+
| ReProver | 229M | 1 | 0/640 |
|
86 |
+
| InternLM2-Step-Prover | 7B | 1 | **5/640** |
|
87 |
+
|
88 |
+
# Citation and Tech Report
|